> We are assuming the longer cf/qual would be written to HDFS billions of > times and would be wasteful. Is that a correct assumption?
Yes, also that's covered a bit in: http://hbase.apache.org/book.html#keysize > > Does the answer change if you use Snappy compression? Any compression will make it better, also have a look at this jira which was recently committed to trunk (adding prefix compression): https://issues.apache.org/jira/browse/HBASE-4218 J-D
